AMD Athlon64 Systems

[Mike A. Harris]

On Mon, 20 Oct 2003, Benny Amorsen wrote:

>> Feel free to debate amongst the nonbelievers however if you 
>> wish...
>
>I haven't encountered any nonbelievers. Everyone knows that 32-bit mode
>is slower than 64-bit mode for the AMD64. The problem was statements
>like this:
>
>"AMD64 architecture provides more benefits to general 64bit apps 
>than broken 64bit sparc processors do.  ;o)"
>
>It is entirely clear that it is the AMD64-architecture that is broken
>for not providing a decent 32-bit mode.

I disagree.  The AMD64 architecture provides a fully x86 
compatible 32bit mode.  The 32bit support is for legacy x86 
application support only, at least that is what AMD is marketing 
it as.

Speaking purely on technical grounds however, I'm under the
assumption that it is technically possible to use some of the new
processor features from within 32bit land.  If you review all
processor features closely, and processor setup, etc. it appears
to me at least that it might be possible, however probably
unsupported by AMD to access enhanced features documented to only 
be useable in 64bit mode from 32bit OS, etc.   That however is 
not fact, but just hypothetical, and based on reading their 
documentation as well as some comments I've read from others.

If it is actually indeed possible, I wouldn't be surprised to see 
games and other things doing nifty hacks to autodetect and use 
the processor's goodies.  I doubt we'd see a new ABI form however 
unless there's significant interest in the hacker community to 
fiddle with it.

If nothing else, it's all interesting in an academic sense.



-- 
Mike A. Harris     ftp://people.redhat.com/mharris
OS Systems Engineer - XFree86 maintainer - Red Hat